Written Responses from RRT Members (PDF: 190KB/22 pages)

Protecting, maintaining and improving the health of all Minnesotans
Date:
June 14, 2010
To:
Provider Peer Grouping Rapid Response Team members
From:
Katie Burns, Health Economics Program
Subject:
Second set of issues for your consideration
Thank you for participating in the Rapid Response Team. In preparation for our second
meeting, I wanted to distribute the attached memos from Mathematica Policy Research,
Inc. They describe the next set of issues for which we would like your input:
1) Should the peer grouping analysis include people without any medical claims in a
given calendar year? If so, what method should be used to attribute those patients to
providers?
2) What specific attribution rules should be used to attribute patients to physician clinics
or medical groups?
We will walk through the memos at our meeting tomorrow afternoon to ensure you have
an opportunity to clarify your understanding of the issues and to ask questions.
Response deadline: We will need your feedback on these issues by Friday, June 25th
at 4:00 pm. Responses may be provided via email to [email protected].
Thanks very much.
Issue #2 – Attribution
MN Hospital Association: Mark Sonneborn
We have differing views from our membership on these issues. Rather than trying
to synthesize, I’ll just give you their responses in their entirety:
1) From [Respondent A]
A. Attribution of Non Users
Support Option 2, using only historical claims data for attributing non-users’ cost and
quality of care in future years as historical data becomes available, and using Option 1
(excluding non-users) in current year.
Option 2 seems more consistent with federal and state reforms to encourage providers to
take greater responsibility for the full spectrum of care for patients they serve and
momentum of providers to organize themselves as Accountable Care Organizations.
Measurement of only users with claims experience perpetuates measuring providers on
an episodic, volume based system and inconsistent with a system based on outcomes and
value which reform efforts are trying to encourage.
Excluding non-users may have perverse incentive for providers to ensure that every
patient in their panel generates at least one claim, necessary or not, on an annual basis in
order to ensure potentially healthier, less resource intensive patients be attributed to
them.
Payers and providers are developing creative payment systems that can fund services that
may not have historically been paid for on a fee for service basis. With continued reliance
on claims as the current best source of utilization and cost data, payers may need to
develop creative processes that allow providers to submit claims or some alternate data
stream for codes and services provided to patients but may not be reimbursed via
historical fee for service but rather through alternative payment methods (e.g. payment
for care management & care coaching).
B. Attribution of Users
Questions:
Assuming unit of cost measure for total cost of care and near total care for chronic
conditions will be on a per member per month basis, how will member months be
proportioned to providers under the multiple-proportional rule?
What do they mean by “managed care patients” under the Attribution to Managed Care
PEs section? Does this section recommend that even if a product requires members to
select a primary care clinic during their enrollment process, the member designated clinic
would be ignored and members would be attributed as if the product were open access?
We would not recommend this methodology, particularly if patient selection and
designation of a medical home would be similarly ignored.
Comments:
Could support Plurality-minimum rule (sole responsibility) or Multiple-proportional rule
(multiple responsible providers proportionately). Even though the plurality-minimum rule
seems more consistent with the primary care focus of ACOs, at this time we support the
Multiple-proportional rule as we believe this option reflects current practice and we will
have much more buy-in from providers.
Perhaps even more useful than the actual provider peer grouping results it will be the
detailed information reflecting a provider’s patient panel pattern of referral care. Many
participants on the PPG Advisory Group commented on the value of the detailed data to
inform providers where their patients seek care; information they are lacking today. If
detailed information for a patient’s entire continuum of care is shared with all providers,
regardless of the attribution rule used, then at least the information to promote better care
coordination would be available to all providers involved with the patient’s care.
The options and recommendations raised in the memo are well supported to produce the
most credible provider peer grouping results but also begins to highlight the importance
to evolve the methodology over time to account for changes in the data source--- claims
data generated from a volume driven health care payment system to data collected from
providers that place greater emphasis on accountability and results at a patient population
level.
2) From [Respondent B]
While I think the vendor has done a thoughtful analysis of the options and their
recommendations, in this case I disagree with them on both fronts.
Concerns about excluding non-users:
It is actually implicit in our goal of improving the value of health care to increase the
number of what show up as non-users in a 12 month claims data set. Health care homes
specifically encourage a system in which providers can deliver services in multiple ways,
many of which won't show up on claim forms. These non-users are a combination of
people who have maintained a sufficient degree of health that has allowed them to stay
out of the health care delivery system and/or have used non-traditional access points,
such as web tools, care manager interactions and phone consults to receive care.
Until we can account for these service types in a way that will allow us to track these
patients, it is critical that we be able to attribute non-users with multi-year data. To do
otherwise encourages overutilization by forcing providers to bring these patients into
their office in order to get "credit"
for them in their data.
I would suggest where available we attribute users with multi-year data, prorating those
who don't have multi-year data initially, and then phase in with multi-year data in the
upcoming analytic rounds as the longitudinal data becomes available. This seems like a
case where expediency as recommended by the vendor is likely to work in opposition to
the care patterns we would like to encourage.
Concerns about attribution recommendation for multi-proportional rule:
The stated preference of Advisory Group was to attribute patients who could reasonably
clearly be attributed and exclude patients for whom attribution was too unclear to
reassure providers that they were being treated fairly.
Because of the very large, comprehensive data set and unit of analysis at the provider
group vs. individual provider level, it was felt that excluding some patients from the
analysis because they could not be reasonably attributed would not weaken the results.
The multi-proportional rule makes more sense when it is critical to assign every patient in
order to maximize the statistical base, which is less of an issue in this case. The Advisory
Group was very comfortable with the idea that not every patient would be attributed.
It was a clear expectation of the legislation to incent primary care providers to coordinate
the care of their patients. Patients seeking care from a hodgepodge of specialists without
involvement of their primary care provider is not something for which the Advisory
Group sought to hold primary care harmless. Using the multi-proportional rule offloads a
significant component of patient costs to specialists which is not the unit of analysis to be
measured in this application and gets primary care docs "off the hook" for coordinating
care. Proportional attribution based on E and M codes creates a disincentive for primary
care follow up and it may even create incentives for patient dumping onto specialists to
keep the cost out of their profile.
The multi-proportional attribution rule does not assure that component costs of care are
distributed appropriately, it just spreads them out. So, for example, if a primary care doc
is carefully managing a patient's diabetes to minimize their ER and inpatient care needs,
and that patient has an unrelated car accident, the primary will still get the preponderance
of the cost of the accident on their results.
I think the intent of the Advisory Group (and certainly mine as facilitator) was to feel as
comfortable as possible defending the validity of the attribution approach to impacted
providers. I think it's a lot easier to get up in front of a group of providers and say that
we excluded 10 or 20% of patients because they couldn't be clearly attributed.
There are other iterations of attribution methodology that could be considered in addition
to the 3 outlined in this memo. It might be most useful if the data team considered how
they might exclude certain patients with multiple primary care provider E and Ms from
different provider entities as an alternative to what is recommended here.
3) [Respondent C]
I agree with the recommendation to not assign patients to a primary care provider if the
patient has not received any health services during the year. I think the other options were
too complex and would therefore be difficult to explain.
I disagree with the recommendation to use the proportional rule for attributing total care
to a provider. A primary care provider should be credited with the total cost of care of a
patient. This will encourage the management of that patient's conditions and will lead to
the best long term outcome for Minnesotans. The methodology that seems the most
appropriate to me is the Plurality-Minimum Rule.
The primary care provider should have responsibility for all of the care of the patient.
They should help direct the patient to more efficient specialists (when they start getting
information about who those providers are). The specialists only have responsibility for
the care associated with the episode they are involved with. Specialists should be
evaluated on the ETG that is attributed to them, not the total cost of care of the patient.
4) [Respondent D]
Should the peer grouping analysis include people without any medical claims in a given
year?
NO! Assignment of non-users when there is not an enrollment process is a guessing
game. That guessing game has the potential to inaccurately measure relative provider
cost performance.
What specific attribution rule should be used to attribute patients to physician clinics or
medical groups?
We support the plurality-minimum rule but with a 50% threshold rather than 30%.
The goal of this process should be to accurately measure relative provider cost
performance.
The goal should not be to attribute as many patients as possible. All of the payer total
cost performance programs that we have been involved with have never had a
threshold lower than 50% including the Medicare Provider Group Practice
demonstration and all of our programs with local health plans.
MN Council of Health Plans: Sue Knudson
Thank you for the opportunity to provide feedback on the ”attribution” issues outlined in
your June 14th letter. The intent of the peer grouping analysis is twofold: a) to inform
consumer choices for high performing care delivery sites based on cost, resource
use/intensity and quality and b) to provide meaningful information to support providers
with performance improvement.
Attribution Questions and MCHP Recommendation:
You requested input on two questions. Members of MCHP conclude the following:
1. Should the peer grouping analysis include people without any medical claims in a
given calendar year? If so, what method should be used to attribute those patients to
providers?
We agree with Mathematica’s recommendation to exclude non-users from current and
future provider peer grouping attribution model. The rationale for this recommendation
is consistent with the issues as outlined in the Mathematica memo.
Our quick and complete agreement with Mathematica’s recommendation should not be
interpreted as a lack of interest in this topic. Just the opposite is true. We appreciate the
efforts put forth by Mathematica to lay out meaningful policy options in this regard.
2. What specific attribution rules should be used to attribute patients to physician clinics
or medical groups?
Mathematica’s recommendation is to use the multiple-proportional rule. Our response to
this question and Mathematica’s recommendation is contingent upon whether meaningful
cohort groups will be used for provider peer grouping. In the absence of clarity
regarding meaningful cohort groups, we provide responses to three scenarios.
a. Primary Care defined as any specialty. This scenario reflects the Advisory
Group’s assumption that, “…the definition of primary care should include any
physician designated as a patient’s primary care physician, regardless of
his/her specialty designation”. No such designation system exists for most
major Minnesota payers, including traditional Medicare.
We would not recommend the multiple-proportional rule be used in this scenario
because it lacks the ability to precisely assign the costs to a truly accountable
provider/visit. It is simply an allocation method across multiple specialties. If
providers are to be grouped with their peers, they will insist that the data should
be pertinent to what they are precisely accountable for (i.e., what they
order/direct/coordinate), and that they be ranked comparatively within a
meaningful cohort. Additionally, downstream economic modeling to understand
the effects of Minnesota’s peer grouping will be confounded by assumptions and
allocations rather than tangible improvements in performance and care-seeking
behavior. In a model where all are accountable, no one is really accountable.
This is further distorted by potentially putting a provider who manages more
conservatively (perhaps via regular monitoring and visits) at risk for another
provider that has fewer visits but is a wasteful manager of care. Negative
outcomes, such as potentially avoidable ED or hospital admissions and higher
costs, would distort the work of the conservative provider.
We believe that comparable cohort groups are critically important.
b. Primary Care as defined by any specialty but limited to those with quality
data.
See our recommendation and rationale in (a) above. We acknowledge that our
concerns are mitigated to the degree that fewer provider specialties are included.
However, we do not yet understand what would be done with costs arising from
the excluded specialties. For example, in the primary care/cardiology/oncology
example raised on page 4 of the Mathematica memo, oncology would not be
tiered due to lack of available quality information. Would the remaining primary
care and cardiology physicians have a portion of the oncology costs attributed to
them, or not? This issue drives one to consider attribution to a single accountable
provider as the Advisory Group originally suggested.
c. Primary Care defined by traditional primary care specialties (FP, IM, Peds,
OB/Gyn)
This is the scenario we favor. We recommend that meaningful provider cohort
groups should be the reliable underpinning on which to attribute people via
administrative claims. For total care provider peer grouping, primary care as
defined by traditional provider specialties such as family practice, internal
medicine, pediatrics and OB/Gyn (recognizing many patients/consumers received
their preventive care from their OB provider) would be a meaningful cohort. This
definition is usable across both primary-care-only and multispecialty practices.
Using a meaningful cohort as the base, the multiple-proportional rule could be
employed as most patients/consumers see only one primary care provider over a
12 month period. However, a more simplified method of majority and most
recent visits could be considered as it would be easier to understand on face value.
As it relates to condition specific provider peer grouping, we support the
recommendation of the Advisory Committee. The committee outlined the
following parameters:
Condition
Diabetes
Who to Measure *
Primary Care
Pneumonia
Heart Failure
Total Knee Replacement
Coronary Artery
Asthma
Endocrinologist
Hospital
Primary Care
Cardiology
Orthopedic
Primary Care
Cardiology
Primary Care
Pediatrician
Pulmonologist
Allergist
* Primary Care should be defined in the manner noted above. Additionally,
cohort groups should be specific to condition, and to who is measured (e.g.
primary care is not comingled with endocrinology for diabetes).
Quality Measurement Attribution
We want to explore a larger concern with you. The issue memo consistently refers to the
attribution methodology being applied to cost and quality measurement. However
clinical quality performance measurement attribution is generally specific to disease, and
it is not as often sourced from administrative claim data alone. Could you please clarify
what type of quality metrics will be derived using these attribution methods? We assume
you are referring to utilization measures. We support using a consistent attribution
methodology across cost and utilization measurement. However, this is another
weakness of the multiple-proportional rule. That rule will not work for utilization
measurement.
Episode Attribution
The previous MDH decision on issue #1 indicated a use of Episode Treatment Groups
(ETG). ETG methodology relies on a different attribution technique that typically
examines the percentage of management and surgery services and uses a threshold (like
25%) to define significant contributors to managing care.
Thank you for the opportunity to comment on the second issue raised by Mathematica.
The members of MCHP look forward to moving the provider peer grouping effort
forward.
MN Medical Association: Janet Silversmith
1. Attribution of Non-Users
The MMA supports an option not explicitly outlined in the Mathematica memo and one
that is generally more consistent with the original recommendation of the Peer Grouping
Advisory Committee. In particular, the MMA does not believe that credible attribution
of non-users is possible without a minimum of three years of claims data. Because the
state does not have access to three years of all-payer claims and enrollment data, the
MMA urges the department not to include non-users in the first two years of analysis;
alternatively, the department could withhold public reporting of results until three years
of data is available, but the MMA is doubtful that is a viable option. When three years of
data are available, the department should (as recommended by the Advisory Committee)
examine the issues of non-use and low-use to understand the impact on results. The
inability to include non-users in the analysis will bias the results. The MMA urges the
department to explicitly identify limitations inherent in the final results, including
inability to attribute non-users, and to caution payers and purchasers in how they use the
results, particularly for payment or network selection purposes as a result of the
limitations.
2. Cost Attribution
The multiple proportional rule does have a variety of advantages over a plan to attribute
all costs to a single provider entity - something that the MMA frequently hears in general
discussions about peer grouping. Physicians feel very strongly about not being held
accountable for care (costs) outside of their control. That said, there are some limitations
associated with this approach that need consideration. First, the notion of "total care"
analysis will be somewhat distorted should this method be implemented. If so, it will be
critical that the department accurately describe what care (and costs) is being measured
so as to avoid confusion (by providers, purchasers, and consumers) as to the actual reach
of the analysis.
As I mentioned during the conference call, the MMA feels very strongly about remaining
true to an evaluation that captures both cost AND quality, as specified in the law. To the
extent that the multiple proportional rule is able to capture additional clinic specialties in
the analysis is meaningless if adequate and representative quality measures are not also
available for that care. Unless the department intends to reexamine what quality
measures will be included in peer grouping, it is unlikely that many single specialty, nonprimary care clinics can be adequately described by the 32 "total care" quality measures.
The MMA urges further review and consideration of this point.
With respect to using E&M visits as the basis for attribution, the MMA is concerned
about cost attribution particularly when non-primary care procedures are included. For
example, an orthopedist (clinic) may only generate 1 E&M consultation visit for a hip
replacement, but it is likely that the cost related to that replacement would
disproportionately fall to the primary care clinic if they represent most of the E&M visits
for that year. The visit to the orthopedist may, or may not, have been recommended or
known to the primary care clinic. A similar situation could also apply for many other
procedures, including hysterectomy, appendectomy, etc. The MMA urges the department
to consider whether proportion of total costs, rather than E&M visits, may be a better
basis for attribution. If E&M visits remain the basis, the MMA strongly encourages the
use of a minimum threshold to reduce inappropriate cost attribution of
procedural/specialty costs to primary care physicians. This recommendation takes into
consideration the open network, open access to care model widely available to
Minnesotans (i.e., little care requires primary care referral/authorization).
Finally, it is unclear to the MMA whether the department is seeking to establish the same
or similar attribution rules for both costs and quality; we look forward to better
understanding the basis for quality attribution in subsequent discussions.
The MMA very much appreciates the opportunity to participate in this process and to
provide comments and input on key methodological issues.
Department of Human Services: Marie Zimmerman
1. Attributing non-users
DHS agrees with Mathematica's recommendation that non-users should be excluded from
the peer-grouping analysis. DHS agrees that if non-users are going to be attributed, we
recommend that option 2 be used and that non-users not be attributed initially in the first
few years until the provider peer grouping dataset has at least 2 years of claims history.
For the perspective of the Medicaid/state public program populations, excluding nonusers is important for two reasons: 1) these populations tend to be more transient and
under utilizers for health care services, therefore this population would be more sensitive
to "false positives" i.e. providers be rewarded for patient they are not treating or
managing; and (2) irresponsible providers sometimes dismiss difficult MA patients which
would inappropriately reward that provider who is not managing their patient.
2. Attribution methodology
DHS does not have a recommendation for an attribution methodology, but would like to
provide some comments. We have a concern about the use of E&M visits as the only
measurement of the PE's level of involvement in a patient's care. This may place an
unfair portion of the patients care on the primary care provider if a specialist only has one
or two E&M visits during the year but makes the decision to provide more costly care,
e.g. surgery, more of that cost could be attributed to the primary care provider. Also, as
health care home (HCH) ramps up, we would recommend that HCH codes/payments be
used to attribute patients as well. This would appropriately reward those HCH who are
efficiently managing a patient's care and have less E&M visits as a result. A specialist
who has more E&M visits with that patient may get rewarded for the HCH effectively
managing that patient's care.
Attribution – Follow up____________________________________________________
Date:
August 5, 2010
To:
Rapid Response Team Members
From:
Katie Burns, Health Economics Program
Subject:
Provider Peer Grouping Issues___________________________________
Thank you for your feedback with respect to the two issues we brought forward for your
consideration. Your thoughtful input caused MDH to consider several interrelated issues
and we now return to you with one methodological decision and to provide greater clarity
around the scope of the provider peer grouping analysis along with some additional
information and request for comment on the different attribution approaches.
Decision on Issue #1: Non-Users
MDH has decided to exclude patients without any medical claims (non-users) in the first
iteration of provider peer grouping. RRT members indicated some differences in opinion
about the desirability of including non-users, but were virtually unanimous in their
opinions that even if we wanted to include them in the first round of peer grouping, we
lack sufficient years of claims data to support credible linkage of these non-users to a
clinic based on prior utilization of a specific provider. MDH leaves open the possibility
of including non-users in future iterations of peer grouping.
Scope of Total Care Analysis
During the June 15, 2010 RRT meeting and in subsequent written comments, RRT
members raised key questions about the scope of the total care physician clinic analysis.
Specifically, the questions centered were related to whether specialists may or may not be
included within the scope of total care peer grouping.
MDH has determined that total care peer grouping will focus on clinics that offer primary
care, whether they are primary care only or are multi-specialty groups that offer primary
care among their other services 1. To be clear, specialists will be included in peer
grouping, but that will occur within condition-specific peer grouping. In addition, the
care provided by specialists will also be within scope of the total care analysis, but the
cost of that care will be attributed back to the clinic responsible for that patient using an
attribution rule yet to be determined. MDH’s larger goal is to preserve the ability to have
one peer group for the total care physician clinic analysis and this approach will provide a
comparable cohort of physician clinics.
1
Over time, as some specialty clinics obtain certification as health care homes, those
clinics will also fall within scope of the total care analysis as they would be expected to
coordinate/provide primary care-like services.
Implications for Attribution Rules
MDH wanted to revisit the attribution rule issue with RRT members as we recognize that
having greater clarity around the scope of the analysis may impact RRT member
feedback on this methodological issue. As we have explored this issue internally and
with our analytical partners at Mathematica, we considered the advantages and
disadvantages of various attribution approaches in greater detail and wanted to share this
information with you as well. The attached memo from Mathematica outlines the
potential attribution rules we shared with you in June with some additional information
about the implications of using them.
Additional Opportunity for Feedback
MDH would appreciate your feedback given the clarification we have provided around
the scope of the total care peer grouping analysis and the implications of various
approaches to attribution. Please send your comments to [email protected] by
4:30 pm on August 16, 2010. Please call Katie Burns at 651.201.3562 with any questions
you may have.
MN Hospital Association: Mark Sonneborn
MHA sent comments from Allina and Park Nicollet
[Respondent A] Response to Provider Peer Grouping Revisit of Attribution Issue
(MDH Rapid Response Team Memo Dated 8/5/2010)
Given MDH’s clarification that total care peer grouping will be focused on peer grouping
primary care groups only, [we are] assuming total costs would be attributed for each
member under each of the different methods as shown in the below example.
Patient John Doe
% of E&M Visits
and Costs
Included in
Attribution?
% Attributed of
Patient’s total cost
under Multiple
Proportional Rule
% Attributed of
Patient’s total cost
under Plurality
Minimum Rule
% Attributed of
Patient’s total cost
under Multiple
Even Rule
Primary Care Clinic 1
Primary Care Clinic 2
29%
25%
Non-Primary
Care Clinics
46%
Yes
Yes
No
54%
46%
0%
0%
0%
0%
0%
0%
0%
Does not meet 30% minimum rule.
Patient excluded otherwise 100%.
Does not meet 30% minimum rule.
Patient excluded otherwise 100%.
Does not meet 30% minimum rule.
Patient excluded, otherwise 100%
[We] supports the Plurality Minimum approach when attributing a patient’s total cost to
primary care clinics. [We] supports assignment to a single clinic with a minimum
threshold because it is most consistent with the momentum of health system reforms
towards a single entity being responsible for the care and coordination of a patient (e.g.,
ACOs, medical homes, etc.). [We] believes sole responsibility assignment to a single
clinic will create the most clarity and greatest incentive for change in the initial years.
While the Multiple Proportional rule has merit at an aggregated cost level since it
recognizes all primary care providers that had an opportunity to coordinate care and
proportionally assigns the total costs to the PCCs, at the detail claims level the multiple
proportional rule raises some practical questions. If Primary Care Clinic 2 requests to see
the detailed claims for the above patient that accounts for the 46% of the patient’s total
costs attributed to the clinic, how will the methodology determine which services are
assigned to Primary Care Clinic 2 versus Primary Care Clinic 1? We would assume all
costs incurred by each clinic would first be attributed to themselves, but the method to
attribute the remaining services is unclear and could create ambiguity and validity
questions for the providers.
[Respondent D]
Three attribution models are being considered:
1.
Multiple proportional rule
2.
Plurality minimum rule
3.
Multiple even rule
As we noted last time, our preferred attribution rule was the plurality minimum rule but
with a 50% threshold because we felt that accurate cost performance measurement was
more important than the numbers of patients attributed.
However, if the use of a 30% threshold is desired in order to increase the patients
numbers in the performance measure, than we would agree that with a 30% threshold, the
multiple proportional rule is the most appropriate.
As [consultant] noted in communication to me on the topic, the important issue on
attribution is that it is transparent and fair.
I support the multiple proportional rule because if the threshold drops to 30%, both the
plurality minimum rule and the multiple even rule result in the groups being 100%
accountable for the cost of care for that patient even though they may have had a very
limited involvement in the cost management given the 30% threshold. The multiple
proportional rule at least somewhat limits how much the attribution of those low
threshold patients impacts the performance of the group.
MN Council of Health Plans: Sue Knudson
Hello Katie,
After reviewing both documents, the MCHP continues to recommend consideration for
the methodology outlined in our original response (attached for your reference). In short,
we could agree with Mathematica's recommendation on attribution in light of MDH's
decision to do total care peer grouping for a meaningful cohort group, primary care only.
However, the alternative and rationale we laid out is still preferred (see top of page 5).
To inform the concerns raised by Mathematica in both the 6/13/10 and 8/6/10 memos and
to validate the methodology selection, we suggest MDH and Mathematica consider
performing an empirical study using the OnPoint data to confirm which is the best
decision. Given how provider groups are organized in MN and how we have observed
consumer care seeking behaviors, Mathematica may see their concerns resolve.
Best Regards,
Sue
MDH Rapid Response Team
Issue #2:
Attribution
Thank you for the opportunity to provide feedback on the ”attribution” issues outlined in
your June 14th letter. The intent of the peer grouping analysis is twofold: a) to inform
consumer choices for high performing care delivery sites based on cost, resource
use/intensity and quality and b) to provide meaningful information to support providers
with performance improvement.
Attribution Questions and MCHP Recommendation:
You requested input on two questions. Members of MCHP conclude the following:
3. Should the peer grouping analysis include people without any medical claims in a
given calendar year? If so, what method should be used to attribute those patients to
providers?
We agree with Mathematica’s recommendation to exclude non-users from current and
future provider peer grouping attribution model. The rationale for this recommendation
is consistent with the issues as outlined in the Mathematica memo.
Our quick and complete agreement with Mathematica’s recommendation should not be
interpreted as a lack of interest in this topic. Just the opposite is true. We appreciate the
efforts put forth by Mathematica to lay out meaningful policy options in this regard.
4. What specific attribution rules should be used to attribute patients to physician clinics
or medical groups?
Mathematica’s recommendation is to use the multiple-proportional rule. Our response to
this question and Mathematica’s recommendation is contingent upon whether meaningful
cohort groups will be used for provider peer grouping. In the absence of clarity
regarding meaningful cohort groups, we provide responses to three scenarios.
a. Primary Care defined as any specialty. This scenario reflects the Advisory
Group’s assumption that, “…the definition of primary care should include any
physician designated as a patient’s primary care physician, regardless of
his/her specialty designation”. No such designation system exists for most
major Minnesota payers, including traditional Medicare.
We would not recommend the multiple-proportional rule be used in this scenario
because it lacks the ability to precisely assign the costs to a truly accountable
provider/visit. It is simply an allocation method across multiple specialties. If
providers are to be grouped with their peers, they will insist that the data should
be pertinent to what they are precisely accountable for (i.e., what they
order/direct/coordinate), and that they be ranked comparatively within a
meaningful cohort. Additionally, downstream economic modeling to understand
the effects of Minnesota’s peer grouping will be confounded by assumptions and
allocations rather than tangible improvements in performance and care-seeking
behavior. In a model where all are accountable, no one is really accountable.
This is further distorted by potentially putting a provider who manages more
conservatively (perhaps via regular monitoring and visits) at risk for another
provider that has fewer visits but is a wasteful manager of care. Negative
outcomes, such as potentially avoidable ED or hospital admissions and higher
costs, would distort the work of the conservative provider.
We believe that comparable cohort groups are critically important.
b. Primary Care as defined by any specialty but limited to those with quality
data.
See our recommendation and rationale in (a) above. We acknowledge that our
concerns are mitigated to the degree that fewer provider specialties are included.
However, we do not yet understand what would be done with costs arising from
the excluded specialties. For example, in the primary care/cardiology/oncology
example raised on page 4 of the Mathematica memo, oncology would not be
tiered due to lack of available quality information. Would the remaining primary
care and cardiology physicians have a portion of the oncology costs attributed to
them, or not? This issue drives one to consider attribution to a single accountable
provider as the Advisory Group originally suggested.
c. Primary Care defined by traditional primary care specialties (FP, IM, Peds,
OB/Gyn)
This is the scenario we favor. We recommend that meaningful provider cohort
groups should be the reliable underpinning on which to attribute people via
administrative claims. For total care provider peer grouping, primary care as
defined by traditional provider specialties such as family practice, internal
medicine, pediatrics and OB/Gyn (recognizing many patients/consumers received
their preventive care from their OB provider) would be a meaningful cohort. This
definition is usable across both primary-care-only and multispecialty practices.
Using a meaningful cohort as the base, the multiple-proportional rule could be
employed as most patients/consumers see only one primary care provider over a
12 month period. However, a more simplified method of majority and most
recent visits could be considered as it would be easier to understand on face value.
As it relates to condition specific provider peer grouping, we support the
recommendation of the Advisory Committee. The committee outlined the
following parameters:
Condition
Diabetes
Pneumonia
Heart Failure
Total Knee Replacement
Coronary Artery
Asthma
Who to Measure *
Primary Care
Endocrinologist
Hospital
Primary Care
Cardiology
Orthopedic
Primary Care
Cardiology
Primary Care
Pediatrician
Pulmonologist
Allergist
* Primary Care should be defined in the manner noted above. Additionally,
cohort groups should be specific to condition, and to who is measured (e.g.
primary care is not comingled with endocrinology for diabetes).
Quality Measurement Attribution
We want to explore a larger concern with you. The issue memo consistently refers to the
attribution methodology being applied to cost and quality measurement. However
clinical quality performance measurement attribution is generally specific to disease, and
it is not as often sourced from administrative claim data alone. Could you please clarify
what type of quality metrics will be derived using these attribution methods? We assume
you are referring to utilization measures. We support using a consistent attribution
methodology across cost and utilization measurement. However, this is another
weakness of the multiple-proportional rule. That rule will not work for utilization
measurement.
Episode Attribution
The previous MDH decision on issue #1 indicated a use of Episode Treatment Groups
(ETG). ETG methodology relies on a different attribution technique that typically
examines the percentage of management and surgery services and uses a threshold (like
25%) to define significant contributors to managing care.
Thank you for the opportunity to comment on the second issue raised by Mathematica.
The members of MCHP look forward to moving the provider peer grouping effort
forward.
MN Medical Association: Janet Silversmith
DATE:
August 16, 2010
TO:
Katie Burns
MN Department of Health
FROM:
Janet Silversmith
RE:
Attribution Rule
The MMA continues to believe that the multiple proportional rule has advantages over a
plan to attribute all costs to a single provider entity. To reiterate previous comments,
there are some limitations associated with this approach that need consideration. First,
the notion of "total care" analysis will be somewhat distorted should this method be
implemented – “total care” costs are no longer being analyzed and attributed, rather only
a specific percentage of total costs are being analyzed and attributed based on the
percentage of E&M visits of the particular primary care clinic. Given the frequent
community discussions about “total cost of care,” the MMA urges the department to
accurately describe what care – and costs – is ultimately going to be measured under peer
grouping so as to avoid confusion – by providers, purchasers, and consumers – as to the
actual reach of the analysis.
With respect to using E&M visits as the basis for attribution given the department’s
decision on primary care clinics,, the MMA finds this approach generally reasonable but
with room for improvement. The MMA remains concerned about cost attribution in
Minnesota’s open access environment. Furthermore, the proposed rule may actually
encourage greater specialist referral by primary care physicians in order for primary care
to decrease their percentage of E&M visits. For example, primary care physicians who
try to handle problems within primary care might have more E&M visits, while other
primary care physicians who refer the patient to specialists more may have fewer E&M
visits yet generate more specialist costs, yet the amount of costs actually attributed could
disproportionately be assigned to the more primary care-oriented group. Finally, the
MMA recognizes the tradeoffs associated with sample size and reliability, but urges
sound methodology to drive the analysis, rather than a need to expand the number of
entities that may be included in the analysis. In other words, if reliability suffers as a
result, then such clinics will need to be dropped from the analysis (note that acceptable
reliability levels have not yet been discussed).
If a percentage of E&M visits remain the basis for attribution of total costs to primary
care clinics, the MMA strongly encourages the use of a minimum threshold to reduce
inappropriate cost attribution of procedural/specialty costs to primary care physicians.
This recommendation takes into consideration the open network, open access-to-care
model widely available to Minnesotans (i.e., little care requires primary care
referral/authorization).
Finally, the MMA urges clarification from the department as to its definition of “primary
care” specialties for purposes of peer grouping attribution.
The MMA very much appreciates the opportunity to participate in this process and to
provide comments.